Mercurial > cpython-lru_cache-weakref
annotate Python/formatter_unicode.c @ 74006:88198b93ff2f
Issue #13560: Add PyUnicode_EncodeLocale()
* Use PyUnicode_EncodeLocale() in time.strftime() if wcsftime() is not
available
* Document my last changes in Misc/NEWS
| author | Victor Stinner <victor.stinner@haypocalc.com> |
|---|---|
| date | Sat, 17 Dec 2011 04:13:41 +0100 |
| parents | 7e37598a25a6 |
| children | 405f65676010 |
| rev | line source |
|---|---|
|
42817
5355337709e1
Implementation of PEP 3101, Advanced String Formatting.
Eric Smith <eric@trueblade.com>
parents:
diff
changeset
|
1 /* implements the unicode (as opposed to string) version of the |
|
5355337709e1
Implementation of PEP 3101, Advanced String Formatting.
Eric Smith <eric@trueblade.com>
parents:
diff
changeset
|
2 built-in formatters for string, int, float. that is, the versions |
|
5355337709e1
Implementation of PEP 3101, Advanced String Formatting.
Eric Smith <eric@trueblade.com>
parents:
diff
changeset
|
3 of int.__float__, etc., that take and return unicode objects */ |
|
5355337709e1
Implementation of PEP 3101, Advanced String Formatting.
Eric Smith <eric@trueblade.com>
parents:
diff
changeset
|
4 |
|
5355337709e1
Implementation of PEP 3101, Advanced String Formatting.
Eric Smith <eric@trueblade.com>
parents:
diff
changeset
|
5 #include "Python.h" |
| 72475 | 6 #include <locale.h> |
| 7 | |
| 8 /* Raises an exception about an unknown presentation type for this | |
| 9 * type. */ | |
| 10 | |
| 11 static void | |
| 12 unknown_presentation_type(Py_UCS4 presentation_type, | |
| 13 const char* type_name) | |
| 14 { | |
| 15 /* %c might be out-of-range, hence the two cases. */ | |
| 16 if (presentation_type > 32 && presentation_type < 128) | |
| 17 PyErr_Format(PyExc_ValueError, | |
| 18 "Unknown format code '%c' " | |
| 19 "for object of type '%.200s'", | |
| 20 (char)presentation_type, | |
| 21 type_name); | |
| 22 else | |
| 23 PyErr_Format(PyExc_ValueError, | |
| 24 "Unknown format code '\\x%x' " | |
| 25 "for object of type '%.200s'", | |
| 26 (unsigned int)presentation_type, | |
| 27 type_name); | |
| 28 } | |
| 29 | |
| 30 static void | |
| 31 invalid_comma_type(Py_UCS4 presentation_type) | |
| 32 { | |
| 33 if (presentation_type > 32 && presentation_type < 128) | |
| 34 PyErr_Format(PyExc_ValueError, | |
| 35 "Cannot specify ',' with '%c'.", | |
| 36 (char)presentation_type); | |
| 37 else | |
| 38 PyErr_Format(PyExc_ValueError, | |
| 39 "Cannot specify ',' with '\\x%x'.", | |
| 40 (unsigned int)presentation_type); | |
| 41 } | |
| 42 | |
| 43 /* | |
| 44 get_integer consumes 0 or more decimal digit characters from an | |
| 45 input string, updates *result with the corresponding positive | |
| 46 integer, and returns the number of digits consumed. | |
| 47 | |
| 48 returns -1 on error. | |
| 49 */ | |
| 50 static int | |
| 51 get_integer(PyObject *str, Py_ssize_t *pos, Py_ssize_t end, | |
| 52 Py_ssize_t *result) | |
| 53 { | |
|
73805
7e37598a25a6
Issue #9530: Fix undefined behaviour due to signed overflow in Python/formatter_unicode.c.
Mark Dickinson <mdickinson@enthought.com>
parents:
72863
diff
changeset
|
54 Py_ssize_t accumulator, digitval; |
| 72475 | 55 int numdigits; |
| 56 accumulator = numdigits = 0; | |
| 57 for (;;(*pos)++, numdigits++) { | |
| 58 if (*pos >= end) | |
| 59 break; | |
| 60 digitval = Py_UNICODE_TODECIMAL(PyUnicode_READ_CHAR(str, *pos)); | |
| 61 if (digitval < 0) | |
| 62 break; | |
| 63 /* | |
|
73805
7e37598a25a6
Issue #9530: Fix undefined behaviour due to signed overflow in Python/formatter_unicode.c.
Mark Dickinson <mdickinson@enthought.com>
parents:
72863
diff
changeset
|
64 Detect possible overflow before it happens: |
|
7e37598a25a6
Issue #9530: Fix undefined behaviour due to signed overflow in Python/formatter_unicode.c.
Mark Dickinson <mdickinson@enthought.com>
parents:
72863
diff
changeset
|
65 |
|
7e37598a25a6
Issue #9530: Fix undefined behaviour due to signed overflow in Python/formatter_unicode.c.
Mark Dickinson <mdickinson@enthought.com>
parents:
72863
diff
changeset
|
66 accumulator * 10 + digitval > PY_SSIZE_T_MAX if and only if |
|
7e37598a25a6
Issue #9530: Fix undefined behaviour due to signed overflow in Python/formatter_unicode.c.
Mark Dickinson <mdickinson@enthought.com>
parents:
72863
diff
changeset
|
67 accumulator > (PY_SSIZE_T_MAX - digitval) / 10. |
| 72475 | 68 */ |
|
73805
7e37598a25a6
Issue #9530: Fix undefined behaviour due to signed overflow in Python/formatter_unicode.c.
Mark Dickinson <mdickinson@enthought.com>
parents:
72863
diff
changeset
|
69 if (accumulator > (PY_SSIZE_T_MAX - digitval) / 10) { |
| 72475 | 70 PyErr_Format(PyExc_ValueError, |
| 71 "Too many decimal digits in format string"); | |
| 72 return -1; | |
| 73 } | |
|
73805
7e37598a25a6
Issue #9530: Fix undefined behaviour due to signed overflow in Python/formatter_unicode.c.
Mark Dickinson <mdickinson@enthought.com>
parents:
72863
diff
changeset
|
74 accumulator = accumulator * 10 + digitval; |
| 72475 | 75 } |
| 76 *result = accumulator; | |
| 77 return numdigits; | |
| 78 } | |
| 79 | |
| 80 /************************************************************************/ | |
| 81 /*********** standard format specifier parsing **************************/ | |
| 82 /************************************************************************/ | |
| 83 | |
| 84 /* returns true if this character is a specifier alignment token */ | |
| 85 Py_LOCAL_INLINE(int) | |
| 86 is_alignment_token(Py_UCS4 c) | |
| 87 { | |
| 88 switch (c) { | |
| 89 case '<': case '>': case '=': case '^': | |
| 90 return 1; | |
| 91 default: | |
| 92 return 0; | |
| 93 } | |
| 94 } | |
| 95 | |
| 96 /* returns true if this character is a sign element */ | |
| 97 Py_LOCAL_INLINE(int) | |
| 98 is_sign_element(Py_UCS4 c) | |
| 99 { | |
| 100 switch (c) { | |
| 101 case ' ': case '+': case '-': | |
| 102 return 1; | |
| 103 default: | |
| 104 return 0; | |
| 105 } | |
| 106 } | |
| 107 | |
| 108 | |
| 109 typedef struct { | |
| 110 Py_UCS4 fill_char; | |
| 111 Py_UCS4 align; | |
| 112 int alternate; | |
| 113 Py_UCS4 sign; | |
| 114 Py_ssize_t width; | |
| 115 int thousands_separators; | |
| 116 Py_ssize_t precision; | |
| 117 Py_UCS4 type; | |
| 118 } InternalFormatSpec; | |
| 119 | |
| 120 #if 0 | |
| 121 /* Occassionally useful for debugging. Should normally be commented out. */ | |
| 122 static void | |
| 123 DEBUG_PRINT_FORMAT_SPEC(InternalFormatSpec *format) | |
| 124 { | |
| 125 printf("internal format spec: fill_char %d\n", format->fill_char); | |
| 126 printf("internal format spec: align %d\n", format->align); | |
| 127 printf("internal format spec: alternate %d\n", format->alternate); | |
| 128 printf("internal format spec: sign %d\n", format->sign); | |
| 129 printf("internal format spec: width %zd\n", format->width); | |
| 130 printf("internal format spec: thousands_separators %d\n", | |
| 131 format->thousands_separators); | |
| 132 printf("internal format spec: precision %zd\n", format->precision); | |
| 133 printf("internal format spec: type %c\n", format->type); | |
| 134 printf("\n"); | |
| 135 } | |
| 136 #endif | |
| 137 | |
| 138 | |
| 139 /* | |
| 140 ptr points to the start of the format_spec, end points just past its end. | |
| 141 fills in format with the parsed information. | |
| 142 returns 1 on success, 0 on failure. | |
| 143 if failure, sets the exception | |
| 144 */ | |
| 145 static int | |
| 146 parse_internal_render_format_spec(PyObject *format_spec, | |
| 147 Py_ssize_t start, Py_ssize_t end, | |
| 148 InternalFormatSpec *format, | |
| 149 char default_type, | |
| 150 char default_align) | |
| 151 { | |
| 152 Py_ssize_t pos = start; | |
| 153 /* end-pos is used throughout this code to specify the length of | |
| 154 the input string */ | |
| 155 #define READ_spec(index) PyUnicode_READ_CHAR(format_spec, index) | |
| 156 | |
| 157 Py_ssize_t consumed; | |
| 158 int align_specified = 0; | |
| 159 | |
| 160 format->fill_char = '\0'; | |
| 161 format->align = default_align; | |
| 162 format->alternate = 0; | |
| 163 format->sign = '\0'; | |
| 164 format->width = -1; | |
| 165 format->thousands_separators = 0; | |
| 166 format->precision = -1; | |
| 167 format->type = default_type; | |
| 168 | |
| 169 /* If the second char is an alignment token, | |
| 170 then parse the fill char */ | |
| 171 if (end-pos >= 2 && is_alignment_token(READ_spec(pos+1))) { | |
| 172 format->align = READ_spec(pos+1); | |
| 173 format->fill_char = READ_spec(pos); | |
| 174 align_specified = 1; | |
| 175 pos += 2; | |
| 176 } | |
| 177 else if (end-pos >= 1 && is_alignment_token(READ_spec(pos))) { | |
| 178 format->align = READ_spec(pos); | |
| 179 align_specified = 1; | |
| 180 ++pos; | |
| 181 } | |
| 182 | |
| 183 /* Parse the various sign options */ | |
| 184 if (end-pos >= 1 && is_sign_element(READ_spec(pos))) { | |
| 185 format->sign = READ_spec(pos); | |
| 186 ++pos; | |
| 187 } | |
| 188 | |
| 189 /* If the next character is #, we're in alternate mode. This only | |
| 190 applies to integers. */ | |
| 191 if (end-pos >= 1 && READ_spec(pos) == '#') { | |
| 192 format->alternate = 1; | |
| 193 ++pos; | |
| 194 } | |
| 195 | |
| 196 /* The special case for 0-padding (backwards compat) */ | |
| 197 if (format->fill_char == '\0' && end-pos >= 1 && READ_spec(pos) == '0') { | |
| 198 format->fill_char = '0'; | |
| 199 if (!align_specified) { | |
| 200 format->align = '='; | |
| 201 } | |
| 202 ++pos; | |
| 203 } | |
| 204 | |
| 205 consumed = get_integer(format_spec, &pos, end, &format->width); | |
| 206 if (consumed == -1) | |
| 207 /* Overflow error. Exception already set. */ | |
| 208 return 0; | |
| 209 | |
| 210 /* If consumed is 0, we didn't consume any characters for the | |
| 211 width. In that case, reset the width to -1, because | |
| 212 get_integer() will have set it to zero. -1 is how we record | |
| 213 that the width wasn't specified. */ | |
| 214 if (consumed == 0) | |
| 215 format->width = -1; | |
| 216 | |
| 217 /* Comma signifies add thousands separators */ | |
| 218 if (end-pos && READ_spec(pos) == ',') { | |
| 219 format->thousands_separators = 1; | |
| 220 ++pos; | |
| 221 } | |
| 222 | |
| 223 /* Parse field precision */ | |
| 224 if (end-pos && READ_spec(pos) == '.') { | |
| 225 ++pos; | |
| 226 | |
| 227 consumed = get_integer(format_spec, &pos, end, &format->precision); | |
| 228 if (consumed == -1) | |
| 229 /* Overflow error. Exception already set. */ | |
| 230 return 0; | |
| 231 | |
| 232 /* Not having a precision after a dot is an error. */ | |
| 233 if (consumed == 0) { | |
| 234 PyErr_Format(PyExc_ValueError, | |
| 235 "Format specifier missing precision"); | |
| 236 return 0; | |
| 237 } | |
| 238 | |
| 239 } | |
| 240 | |
| 241 /* Finally, parse the type field. */ | |
| 242 | |
| 243 if (end-pos > 1) { | |
| 244 /* More than one char remain, invalid conversion spec. */ | |
| 245 PyErr_Format(PyExc_ValueError, "Invalid conversion specification"); | |
| 246 return 0; | |
| 247 } | |
| 248 | |
| 249 if (end-pos == 1) { | |
| 250 format->type = READ_spec(pos); | |
| 251 ++pos; | |
| 252 } | |
| 253 | |
| 254 /* Do as much validating as we can, just by looking at the format | |
| 255 specifier. Do not take into account what type of formatting | |
| 256 we're doing (int, float, string). */ | |
| 257 | |
| 258 if (format->thousands_separators) { | |
| 259 switch (format->type) { | |
| 260 case 'd': | |
| 261 case 'e': | |
| 262 case 'f': | |
| 263 case 'g': | |
| 264 case 'E': | |
| 265 case 'G': | |
| 266 case '%': | |
| 267 case 'F': | |
| 268 case '\0': | |
| 269 /* These are allowed. See PEP 378.*/ | |
| 270 break; | |
| 271 default: | |
| 272 invalid_comma_type(format->type); | |
| 273 return 0; | |
| 274 } | |
| 275 } | |
| 276 | |
| 277 if (format->fill_char > 127 || format->align > 127 || | |
| 278 format->sign > 127) { | |
| 279 PyErr_SetString(PyExc_ValueError, "fill character too large"); | |
| 280 return 0; | |
| 281 } | |
| 282 | |
| 283 return 1; | |
| 284 } | |
| 285 | |
| 286 /* Calculate the padding needed. */ | |
| 287 static void | |
| 288 calc_padding(Py_ssize_t nchars, Py_ssize_t width, Py_UCS4 align, | |
| 289 Py_ssize_t *n_lpadding, Py_ssize_t *n_rpadding, | |
| 290 Py_ssize_t *n_total) | |
| 291 { | |
| 292 if (width >= 0) { | |
| 293 if (nchars > width) | |
| 294 *n_total = nchars; | |
| 295 else | |
| 296 *n_total = width; | |
| 297 } | |
| 298 else { | |
| 299 /* not specified, use all of the chars and no more */ | |
| 300 *n_total = nchars; | |
| 301 } | |
| 302 | |
| 303 /* Figure out how much leading space we need, based on the | |
| 304 aligning */ | |
| 305 if (align == '>') | |
| 306 *n_lpadding = *n_total - nchars; | |
| 307 else if (align == '^') | |
| 308 *n_lpadding = (*n_total - nchars) / 2; | |
| 309 else if (align == '<' || align == '=') | |
| 310 *n_lpadding = 0; | |
| 311 else { | |
| 312 /* We should never have an unspecified alignment. */ | |
| 313 *n_lpadding = 0; | |
| 314 assert(0); | |
| 315 } | |
| 316 | |
| 317 *n_rpadding = *n_total - nchars - *n_lpadding; | |
| 318 } | |
| 319 | |
| 320 static void | |
| 321 unicode_fill(PyObject *str, Py_ssize_t start, Py_ssize_t end, Py_UCS4 ch) | |
| 322 { | |
| 323 int kind = PyUnicode_KIND(str); | |
| 324 void *data = PyUnicode_DATA(str); | |
| 325 while (start < end) | |
| 326 PyUnicode_WRITE(kind, data, start++, ch); | |
| 327 } | |
| 328 | |
| 329 /* Do the padding, and return a pointer to where the caller-supplied | |
| 330 content goes. */ | |
| 331 static Py_ssize_t | |
| 332 fill_padding(PyObject *s, Py_ssize_t start, Py_ssize_t nchars, | |
| 333 Py_UCS4 fill_char, Py_ssize_t n_lpadding, | |
| 334 Py_ssize_t n_rpadding) | |
| 335 { | |
| 336 /* Pad on left. */ | |
| 337 if (n_lpadding) | |
| 338 unicode_fill(s, start, start + n_lpadding, fill_char); | |
| 339 | |
| 340 /* Pad on right. */ | |
| 341 if (n_rpadding) | |
|
72495
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
342 unicode_fill(s, start + nchars + n_lpadding, |
| 72475 | 343 start + nchars + n_lpadding + n_rpadding, fill_char); |
| 344 | |
| 345 /* Pointer to the user content. */ | |
| 346 return start + n_lpadding; | |
| 347 } | |
| 348 | |
| 349 /************************************************************************/ | |
| 350 /*********** common routines for numeric formatting *********************/ | |
| 351 /************************************************************************/ | |
| 352 | |
| 353 /* Locale type codes. */ | |
| 354 #define LT_CURRENT_LOCALE 0 | |
| 355 #define LT_DEFAULT_LOCALE 1 | |
| 356 #define LT_NO_LOCALE 2 | |
| 357 | |
| 358 /* Locale info needed for formatting integers and the part of floats | |
| 359 before and including the decimal. Note that locales only support | |
| 360 8-bit chars, not unicode. */ | |
| 361 typedef struct { | |
| 362 char *decimal_point; | |
| 363 char *thousands_sep; | |
| 364 char *grouping; | |
| 365 } LocaleInfo; | |
| 366 | |
| 367 /* describes the layout for an integer, see the comment in | |
| 368 calc_number_widths() for details */ | |
| 369 typedef struct { | |
| 370 Py_ssize_t n_lpadding; | |
| 371 Py_ssize_t n_prefix; | |
| 372 Py_ssize_t n_spadding; | |
| 373 Py_ssize_t n_rpadding; | |
| 374 char sign; | |
| 375 Py_ssize_t n_sign; /* number of digits needed for sign (0/1) */ | |
| 376 Py_ssize_t n_grouped_digits; /* Space taken up by the digits, including | |
| 377 any grouping chars. */ | |
| 378 Py_ssize_t n_decimal; /* 0 if only an integer */ | |
| 379 Py_ssize_t n_remainder; /* Digits in decimal and/or exponent part, | |
| 380 excluding the decimal itself, if | |
| 381 present. */ | |
| 382 | |
| 383 /* These 2 are not the widths of fields, but are needed by | |
| 384 STRINGLIB_GROUPING. */ | |
| 385 Py_ssize_t n_digits; /* The number of digits before a decimal | |
| 386 or exponent. */ | |
| 387 Py_ssize_t n_min_width; /* The min_width we used when we computed | |
| 388 the n_grouped_digits width. */ | |
| 389 } NumberFieldWidths; | |
|
42817
5355337709e1
Implementation of PEP 3101, Advanced String Formatting.
Eric Smith <eric@trueblade.com>
parents:
diff
changeset
|
390 |
|
47393
0da27fdf0192
Refactor and clean up str.format() code (and helpers) in advance of optimizations.
Eric Smith <eric@trueblade.com>
parents:
42817
diff
changeset
|
391 |
| 72475 | 392 /* Given a number of the form: |
| 393 digits[remainder] | |
| 394 where ptr points to the start and end points to the end, find where | |
| 395 the integer part ends. This could be a decimal, an exponent, both, | |
| 396 or neither. | |
| 397 If a decimal point is present, set *has_decimal and increment | |
| 398 remainder beyond it. | |
| 399 Results are undefined (but shouldn't crash) for improperly | |
| 400 formatted strings. | |
| 401 */ | |
| 402 static void | |
| 403 parse_number(PyObject *s, Py_ssize_t pos, Py_ssize_t end, | |
| 404 Py_ssize_t *n_remainder, int *has_decimal) | |
| 405 { | |
| 406 Py_ssize_t remainder; | |
| 407 | |
| 408 while (pos<end && isdigit(PyUnicode_READ_CHAR(s, pos))) | |
| 409 ++pos; | |
| 410 remainder = pos; | |
| 411 | |
| 412 /* Does remainder start with a decimal point? */ | |
| 413 *has_decimal = pos<end && PyUnicode_READ_CHAR(s, remainder) == '.'; | |
| 414 | |
| 415 /* Skip the decimal point. */ | |
| 416 if (*has_decimal) | |
| 417 remainder++; | |
| 418 | |
| 419 *n_remainder = end - remainder; | |
| 420 } | |
| 421 | |
| 422 /* not all fields of format are used. for example, precision is | |
| 423 unused. should this take discrete params in order to be more clear | |
| 424 about what it does? or is passing a single format parameter easier | |
| 425 and more efficient enough to justify a little obfuscation? */ | |
| 426 static Py_ssize_t | |
| 427 calc_number_widths(NumberFieldWidths *spec, Py_ssize_t n_prefix, | |
| 428 Py_UCS4 sign_char, PyObject *number, Py_ssize_t n_start, | |
| 429 Py_ssize_t n_end, Py_ssize_t n_remainder, | |
| 430 int has_decimal, const LocaleInfo *locale, | |
| 431 const InternalFormatSpec *format) | |
| 432 { | |
| 433 Py_ssize_t n_non_digit_non_padding; | |
| 434 Py_ssize_t n_padding; | |
| 435 | |
| 436 spec->n_digits = n_end - n_start - n_remainder - (has_decimal?1:0); | |
| 437 spec->n_lpadding = 0; | |
| 438 spec->n_prefix = n_prefix; | |
| 439 spec->n_decimal = has_decimal ? strlen(locale->decimal_point) : 0; | |
| 440 spec->n_remainder = n_remainder; | |
| 441 spec->n_spadding = 0; | |
| 442 spec->n_rpadding = 0; | |
| 443 spec->sign = '\0'; | |
| 444 spec->n_sign = 0; | |
| 445 | |
| 446 /* the output will look like: | |
| 447 | | | |
| 448 | <lpadding> <sign> <prefix> <spadding> <grouped_digits> <decimal> <remainder> <rpadding> | | |
| 449 | | | |
| 450 | |
| 451 sign is computed from format->sign and the actual | |
| 452 sign of the number | |
| 453 | |
| 454 prefix is given (it's for the '0x' prefix) | |
| 455 | |
| 456 digits is already known | |
| 457 | |
| 458 the total width is either given, or computed from the | |
| 459 actual digits | |
| 460 | |
| 461 only one of lpadding, spadding, and rpadding can be non-zero, | |
| 462 and it's calculated from the width and other fields | |
| 463 */ | |
| 464 | |
| 465 /* compute the various parts we're going to write */ | |
| 466 switch (format->sign) { | |
| 467 case '+': | |
| 468 /* always put a + or - */ | |
| 469 spec->n_sign = 1; | |
| 470 spec->sign = (sign_char == '-' ? '-' : '+'); | |
| 471 break; | |
| 472 case ' ': | |
| 473 spec->n_sign = 1; | |
| 474 spec->sign = (sign_char == '-' ? '-' : ' '); | |
| 475 break; | |
| 476 default: | |
| 477 /* Not specified, or the default (-) */ | |
| 478 if (sign_char == '-') { | |
| 479 spec->n_sign = 1; | |
| 480 spec->sign = '-'; | |
| 481 } | |
| 482 } | |
| 483 | |
| 484 /* The number of chars used for non-digits and non-padding. */ | |
| 485 n_non_digit_non_padding = spec->n_sign + spec->n_prefix + spec->n_decimal + | |
| 486 spec->n_remainder; | |
| 487 | |
| 488 /* min_width can go negative, that's okay. format->width == -1 means | |
| 489 we don't care. */ | |
| 490 if (format->fill_char == '0' && format->align == '=') | |
| 491 spec->n_min_width = format->width - n_non_digit_non_padding; | |
| 492 else | |
| 493 spec->n_min_width = 0; | |
| 494 | |
| 495 if (spec->n_digits == 0) | |
| 496 /* This case only occurs when using 'c' formatting, we need | |
| 497 to special case it because the grouping code always wants | |
| 498 to have at least one character. */ | |
| 499 spec->n_grouped_digits = 0; | |
| 500 else | |
| 501 spec->n_grouped_digits = _PyUnicode_InsertThousandsGrouping( | |
|
72705
05ed6e5f2cf4
Add asciilib: similar to ucs1, ucs2 and ucs4 library, but specialized to ASCII
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72497
diff
changeset
|
502 NULL, PyUnicode_1BYTE_KIND, NULL, 0, NULL, |
| 72475 | 503 spec->n_digits, spec->n_min_width, |
| 504 locale->grouping, locale->thousands_sep); | |
| 505 | |
| 506 /* Given the desired width and the total of digit and non-digit | |
| 507 space we consume, see if we need any padding. format->width can | |
| 508 be negative (meaning no padding), but this code still works in | |
| 509 that case. */ | |
| 510 n_padding = format->width - | |
| 511 (n_non_digit_non_padding + spec->n_grouped_digits); | |
| 512 if (n_padding > 0) { | |
| 513 /* Some padding is needed. Determine if it's left, space, or right. */ | |
| 514 switch (format->align) { | |
| 515 case '<': | |
| 516 spec->n_rpadding = n_padding; | |
| 517 break; | |
| 518 case '^': | |
| 519 spec->n_lpadding = n_padding / 2; | |
| 520 spec->n_rpadding = n_padding - spec->n_lpadding; | |
| 521 break; | |
| 522 case '=': | |
| 523 spec->n_spadding = n_padding; | |
| 524 break; | |
| 525 case '>': | |
| 526 spec->n_lpadding = n_padding; | |
| 527 break; | |
| 528 default: | |
| 529 /* Shouldn't get here, but treat it as '>' */ | |
| 530 spec->n_lpadding = n_padding; | |
| 531 assert(0); | |
| 532 break; | |
| 533 } | |
| 534 } | |
| 535 return spec->n_lpadding + spec->n_sign + spec->n_prefix + | |
| 536 spec->n_spadding + spec->n_grouped_digits + spec->n_decimal + | |
| 537 spec->n_remainder + spec->n_rpadding; | |
| 538 } | |
| 539 | |
| 540 /* Fill in the digit parts of a numbers's string representation, | |
| 541 as determined in calc_number_widths(). | |
|
72495
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
542 Return -1 on error, or 0 on success. */ |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
543 static int |
| 72475 | 544 fill_number(PyObject *out, Py_ssize_t pos, const NumberFieldWidths *spec, |
| 545 PyObject *digits, Py_ssize_t d_start, Py_ssize_t d_end, | |
|
72495
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
546 PyObject *prefix, Py_ssize_t p_start, |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
547 Py_UCS4 fill_char, |
| 72475 | 548 LocaleInfo *locale, int toupper) |
| 549 { | |
| 550 /* Used to keep track of digits, decimal, and remainder. */ | |
| 551 Py_ssize_t d_pos = d_start; | |
| 552 unsigned int kind = PyUnicode_KIND(out); | |
| 553 void *data = PyUnicode_DATA(out); | |
| 554 | |
| 555 #ifndef NDEBUG | |
| 556 Py_ssize_t r; | |
| 557 #endif | |
| 558 | |
| 559 if (spec->n_lpadding) { | |
| 560 unicode_fill(out, pos, pos + spec->n_lpadding, fill_char); | |
| 561 pos += spec->n_lpadding; | |
| 562 } | |
| 563 if (spec->n_sign == 1) { | |
| 564 PyUnicode_WRITE(kind, data, pos++, spec->sign); | |
| 565 } | |
| 566 if (spec->n_prefix) { | |
|
72497
f05c2df39ea8
fill_number() and format_string_internal() check for PyUnicode_CopyCharacters() failure
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72496
diff
changeset
|
567 if (PyUnicode_CopyCharacters(out, pos, |
|
f05c2df39ea8
fill_number() and format_string_internal() check for PyUnicode_CopyCharacters() failure
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72496
diff
changeset
|
568 prefix, p_start, |
|
f05c2df39ea8
fill_number() and format_string_internal() check for PyUnicode_CopyCharacters() failure
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72496
diff
changeset
|
569 spec->n_prefix) < 0) |
|
f05c2df39ea8
fill_number() and format_string_internal() check for PyUnicode_CopyCharacters() failure
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72496
diff
changeset
|
570 return -1; |
| 72475 | 571 if (toupper) { |
| 572 Py_ssize_t t; | |
| 573 /* XXX if the upper-case prefix is wider than the target | |
| 574 buffer, the caller should have allocated a wider string, | |
| 575 but currently doesn't. */ | |
| 576 for (t = 0; t < spec->n_prefix; ++t) | |
| 577 PyUnicode_WRITE(kind, data, pos + t, | |
| 578 Py_UNICODE_TOUPPER( | |
| 579 PyUnicode_READ(kind, data, pos + t))); | |
| 580 } | |
| 581 pos += spec->n_prefix; | |
| 582 } | |
| 583 if (spec->n_spadding) { | |
| 584 unicode_fill(out, pos, pos + spec->n_spadding, fill_char); | |
| 585 pos += spec->n_spadding; | |
| 586 } | |
| 587 | |
| 588 /* Only for type 'c' special case, it has no digits. */ | |
| 589 if (spec->n_digits != 0) { | |
| 590 /* Fill the digits with InsertThousandsGrouping. */ | |
|
72496
ad337be523bf
fill_number() ensures that the 'digits' string is ready
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72495
diff
changeset
|
591 char *pdigits; |
|
ad337be523bf
fill_number() ensures that the 'digits' string is ready
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72495
diff
changeset
|
592 if (PyUnicode_READY(digits)) |
|
ad337be523bf
fill_number() ensures that the 'digits' string is ready
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72495
diff
changeset
|
593 return -1; |
|
ad337be523bf
fill_number() ensures that the 'digits' string is ready
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72495
diff
changeset
|
594 pdigits = PyUnicode_DATA(digits); |
| 72475 | 595 if (PyUnicode_KIND(digits) < kind) { |
| 596 pdigits = _PyUnicode_AsKind(digits, kind); | |
|
72495
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
597 if (pdigits == NULL) |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
598 return -1; |
| 72475 | 599 } |
| 600 #ifndef NDEBUG | |
| 601 r = | |
| 602 #endif | |
| 603 _PyUnicode_InsertThousandsGrouping( | |
|
72705
05ed6e5f2cf4
Add asciilib: similar to ucs1, ucs2 and ucs4 library, but specialized to ASCII
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72497
diff
changeset
|
604 out, kind, |
|
72796
c25262e97304
Change PyUnicode_KIND to 1,2,4. Drop _KIND_SIZE and _CHARACTER_SIZE.
Martin v. Löwis <martin@v.loewis.de>
parents:
72714
diff
changeset
|
605 (char*)data + kind * pos, |
| 72475 | 606 spec->n_grouped_digits, |
|
72796
c25262e97304
Change PyUnicode_KIND to 1,2,4. Drop _KIND_SIZE and _CHARACTER_SIZE.
Martin v. Löwis <martin@v.loewis.de>
parents:
72714
diff
changeset
|
607 pdigits + kind * d_pos, |
| 72475 | 608 spec->n_digits, spec->n_min_width, |
| 609 locale->grouping, locale->thousands_sep); | |
| 610 #ifndef NDEBUG | |
| 611 assert(r == spec->n_grouped_digits); | |
| 612 #endif | |
| 613 if (PyUnicode_KIND(digits) < kind) | |
| 614 PyMem_Free(pdigits); | |
| 615 d_pos += spec->n_digits; | |
| 616 } | |
| 617 if (toupper) { | |
| 618 Py_ssize_t t; | |
| 619 for (t = 0; t < spec->n_grouped_digits; ++t) | |
| 620 PyUnicode_WRITE(kind, data, pos + t, | |
| 621 Py_UNICODE_TOUPPER( | |
| 622 PyUnicode_READ(kind, data, pos + t))); | |
| 623 } | |
| 624 pos += spec->n_grouped_digits; | |
| 625 | |
| 626 if (spec->n_decimal) { | |
| 627 Py_ssize_t t; | |
| 628 for (t = 0; t < spec->n_decimal; ++t) | |
| 629 PyUnicode_WRITE(kind, data, pos + t, | |
| 630 locale->decimal_point[t]); | |
| 631 pos += spec->n_decimal; | |
| 632 d_pos += 1; | |
| 633 } | |
| 634 | |
| 635 if (spec->n_remainder) { | |
|
72497
f05c2df39ea8
fill_number() and format_string_internal() check for PyUnicode_CopyCharacters() failure
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72496
diff
changeset
|
636 if (PyUnicode_CopyCharacters(out, pos, digits, d_pos, spec->n_remainder) < 0) |
|
f05c2df39ea8
fill_number() and format_string_internal() check for PyUnicode_CopyCharacters() failure
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72496
diff
changeset
|
637 return -1; |
| 72475 | 638 pos += spec->n_remainder; |
| 639 d_pos += spec->n_remainder; | |
| 640 } | |
| 641 | |
| 642 if (spec->n_rpadding) { | |
| 643 unicode_fill(out, pos, pos + spec->n_rpadding, fill_char); | |
| 644 pos += spec->n_rpadding; | |
| 645 } | |
|
72495
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
646 return 0; |
| 72475 | 647 } |
| 648 | |
| 649 static char no_grouping[1] = {CHAR_MAX}; | |
| 650 | |
| 651 /* Find the decimal point character(s?), thousands_separator(s?), and | |
| 652 grouping description, either for the current locale if type is | |
| 653 LT_CURRENT_LOCALE, a hard-coded locale if LT_DEFAULT_LOCALE, or | |
| 654 none if LT_NO_LOCALE. */ | |
| 655 static void | |
| 656 get_locale_info(int type, LocaleInfo *locale_info) | |
| 657 { | |
| 658 switch (type) { | |
| 659 case LT_CURRENT_LOCALE: { | |
| 660 struct lconv *locale_data = localeconv(); | |
| 661 locale_info->decimal_point = locale_data->decimal_point; | |
| 662 locale_info->thousands_sep = locale_data->thousands_sep; | |
| 663 locale_info->grouping = locale_data->grouping; | |
| 664 break; | |
| 665 } | |
| 666 case LT_DEFAULT_LOCALE: | |
| 667 locale_info->decimal_point = "."; | |
| 668 locale_info->thousands_sep = ","; | |
| 669 locale_info->grouping = "\3"; /* Group every 3 characters. The | |
| 670 (implicit) trailing 0 means repeat | |
| 671 infinitely. */ | |
| 672 break; | |
| 673 case LT_NO_LOCALE: | |
| 674 locale_info->decimal_point = "."; | |
| 675 locale_info->thousands_sep = ""; | |
| 676 locale_info->grouping = no_grouping; | |
| 677 break; | |
| 678 default: | |
| 679 assert(0); | |
| 680 } | |
| 681 } | |
| 682 | |
| 683 /************************************************************************/ | |
| 684 /*********** string formatting ******************************************/ | |
| 685 /************************************************************************/ | |
| 686 | |
| 687 static PyObject * | |
| 688 format_string_internal(PyObject *value, const InternalFormatSpec *format) | |
| 689 { | |
| 690 Py_ssize_t lpad; | |
| 691 Py_ssize_t rpad; | |
| 692 Py_ssize_t total; | |
| 693 Py_ssize_t pos; | |
|
72863
afa42c04f0a3
Fix misuse of PyUnicode_GET_SIZE, use PyUnicode_GET_LENGTH instead
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72796
diff
changeset
|
694 Py_ssize_t len = PyUnicode_GET_LENGTH(value); |
| 72475 | 695 PyObject *result = NULL; |
| 696 int maxchar = 127; | |
| 697 | |
| 698 /* sign is not allowed on strings */ | |
| 699 if (format->sign != '\0') { | |
| 700 PyErr_SetString(PyExc_ValueError, | |
| 701 "Sign not allowed in string format specifier"); | |
| 702 goto done; | |
| 703 } | |
| 704 | |
| 705 /* alternate is not allowed on strings */ | |
| 706 if (format->alternate) { | |
| 707 PyErr_SetString(PyExc_ValueError, | |
| 708 "Alternate form (#) not allowed in string format " | |
| 709 "specifier"); | |
| 710 goto done; | |
| 711 } | |
| 712 | |
| 713 /* '=' alignment not allowed on strings */ | |
| 714 if (format->align == '=') { | |
| 715 PyErr_SetString(PyExc_ValueError, | |
| 716 "'=' alignment not allowed " | |
| 717 "in string format specifier"); | |
| 718 goto done; | |
| 719 } | |
| 720 | |
| 721 /* if precision is specified, output no more that format.precision | |
| 722 characters */ | |
| 723 if (format->precision >= 0 && len >= format->precision) { | |
| 724 len = format->precision; | |
| 725 } | |
| 726 | |
| 727 calc_padding(len, format->width, format->align, &lpad, &rpad, &total); | |
| 728 | |
| 729 /* allocate the resulting string */ | |
| 730 result = PyUnicode_New(total, maxchar); | |
| 731 if (result == NULL) | |
| 732 goto done; | |
| 733 | |
| 734 /* Write into that space. First the padding. */ | |
| 735 pos = fill_padding(result, 0, len, | |
| 736 format->fill_char=='\0'?' ':format->fill_char, | |
| 737 lpad, rpad); | |
| 738 | |
| 739 /* Then the source string. */ | |
|
72497
f05c2df39ea8
fill_number() and format_string_internal() check for PyUnicode_CopyCharacters() failure
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72496
diff
changeset
|
740 if (PyUnicode_CopyCharacters(result, pos, value, 0, len) < 0) |
|
f05c2df39ea8
fill_number() and format_string_internal() check for PyUnicode_CopyCharacters() failure
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72496
diff
changeset
|
741 Py_CLEAR(result); |
| 72475 | 742 |
| 743 done: | |
| 744 return result; | |
| 745 } | |
| 746 | |
| 747 | |
| 748 /************************************************************************/ | |
| 749 /*********** long formatting ********************************************/ | |
| 750 /************************************************************************/ | |
| 751 | |
| 752 typedef PyObject* | |
| 753 (*IntOrLongToString)(PyObject *value, int base); | |
| 754 | |
| 755 static PyObject * | |
| 756 format_int_or_long_internal(PyObject *value, const InternalFormatSpec *format, | |
| 757 IntOrLongToString tostring) | |
| 758 { | |
| 759 PyObject *result = NULL; | |
| 760 int maxchar = 127; | |
| 761 PyObject *tmp = NULL; | |
| 762 Py_ssize_t inumeric_chars; | |
| 763 Py_UCS4 sign_char = '\0'; | |
| 764 Py_ssize_t n_digits; /* count of digits need from the computed | |
| 765 string */ | |
| 766 Py_ssize_t n_remainder = 0; /* Used only for 'c' formatting, which | |
| 767 produces non-digits */ | |
| 768 Py_ssize_t n_prefix = 0; /* Count of prefix chars, (e.g., '0x') */ | |
| 769 Py_ssize_t n_total; | |
| 770 Py_ssize_t prefix; | |
| 771 NumberFieldWidths spec; | |
| 772 long x; | |
|
72495
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
773 int err; |
| 72475 | 774 |
| 775 /* Locale settings, either from the actual locale or | |
| 776 from a hard-code pseudo-locale */ | |
| 777 LocaleInfo locale; | |
| 778 | |
| 779 /* no precision allowed on integers */ | |
| 780 if (format->precision != -1) { | |
| 781 PyErr_SetString(PyExc_ValueError, | |
| 782 "Precision not allowed in integer format specifier"); | |
| 783 goto done; | |
| 784 } | |
| 785 | |
| 786 /* special case for character formatting */ | |
| 787 if (format->type == 'c') { | |
| 788 /* error to specify a sign */ | |
| 789 if (format->sign != '\0') { | |
| 790 PyErr_SetString(PyExc_ValueError, | |
| 791 "Sign not allowed with integer" | |
| 792 " format specifier 'c'"); | |
| 793 goto done; | |
| 794 } | |
| 795 | |
| 796 /* taken from unicodeobject.c formatchar() */ | |
| 797 /* Integer input truncated to a character */ | |
| 798 /* XXX: won't work for int */ | |
| 799 x = PyLong_AsLong(value); | |
| 800 if (x == -1 && PyErr_Occurred()) | |
| 801 goto done; | |
| 802 if (x < 0 || x > 0x10ffff) { | |
| 803 PyErr_SetString(PyExc_OverflowError, | |
| 804 "%c arg not in range(0x110000) " | |
| 805 "(wide Python build)"); | |
| 806 goto done; | |
| 807 } | |
| 808 tmp = PyUnicode_FromOrdinal(x); | |
| 809 inumeric_chars = 0; | |
| 810 n_digits = 1; | |
| 811 if (x > maxchar) | |
| 812 maxchar = x; | |
| 813 | |
| 814 /* As a sort-of hack, we tell calc_number_widths that we only | |
| 815 have "remainder" characters. calc_number_widths thinks | |
| 816 these are characters that don't get formatted, only copied | |
| 817 into the output string. We do this for 'c' formatting, | |
| 818 because the characters are likely to be non-digits. */ | |
| 819 n_remainder = 1; | |
| 820 } | |
| 821 else { | |
| 822 int base; | |
| 823 int leading_chars_to_skip = 0; /* Number of characters added by | |
| 824 PyNumber_ToBase that we want to | |
| 825 skip over. */ | |
| 826 | |
| 827 /* Compute the base and how many characters will be added by | |
| 828 PyNumber_ToBase */ | |
| 829 switch (format->type) { | |
| 830 case 'b': | |
| 831 base = 2; | |
| 832 leading_chars_to_skip = 2; /* 0b */ | |
| 833 break; | |
| 834 case 'o': | |
| 835 base = 8; | |
| 836 leading_chars_to_skip = 2; /* 0o */ | |
| 837 break; | |
| 838 case 'x': | |
| 839 case 'X': | |
| 840 base = 16; | |
| 841 leading_chars_to_skip = 2; /* 0x */ | |
| 842 break; | |
| 843 default: /* shouldn't be needed, but stops a compiler warning */ | |
| 844 case 'd': | |
| 845 case 'n': | |
| 846 base = 10; | |
| 847 break; | |
| 848 } | |
| 849 | |
| 850 /* The number of prefix chars is the same as the leading | |
| 851 chars to skip */ | |
| 852 if (format->alternate) | |
| 853 n_prefix = leading_chars_to_skip; | |
| 854 | |
| 855 /* Do the hard part, converting to a string in a given base */ | |
| 856 tmp = tostring(value, base); | |
| 857 if (tmp == NULL || PyUnicode_READY(tmp) == -1) | |
| 858 goto done; | |
| 859 | |
| 860 inumeric_chars = 0; | |
| 861 n_digits = PyUnicode_GET_LENGTH(tmp); | |
| 862 | |
| 863 prefix = inumeric_chars; | |
| 864 | |
| 865 /* Is a sign character present in the output? If so, remember it | |
| 866 and skip it */ | |
| 867 if (PyUnicode_READ_CHAR(tmp, inumeric_chars) == '-') { | |
| 868 sign_char = '-'; | |
| 869 ++prefix; | |
| 870 ++leading_chars_to_skip; | |
| 871 } | |
| 872 | |
| 873 /* Skip over the leading chars (0x, 0b, etc.) */ | |
| 874 n_digits -= leading_chars_to_skip; | |
| 875 inumeric_chars += leading_chars_to_skip; | |
| 876 } | |
| 877 | |
| 878 /* Determine the grouping, separator, and decimal point, if any. */ | |
| 879 get_locale_info(format->type == 'n' ? LT_CURRENT_LOCALE : | |
| 880 (format->thousands_separators ? | |
| 881 LT_DEFAULT_LOCALE : | |
| 882 LT_NO_LOCALE), | |
| 883 &locale); | |
| 884 | |
| 885 /* Calculate how much memory we'll need. */ | |
| 886 n_total = calc_number_widths(&spec, n_prefix, sign_char, tmp, inumeric_chars, | |
| 887 inumeric_chars + n_digits, n_remainder, 0, &locale, format); | |
| 888 | |
| 889 /* Allocate the memory. */ | |
| 890 result = PyUnicode_New(n_total, maxchar); | |
| 891 if (!result) | |
| 892 goto done; | |
| 893 | |
| 894 /* Populate the memory. */ | |
|
72495
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
895 err = fill_number(result, 0, &spec, |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
896 tmp, inumeric_chars, inumeric_chars + n_digits, |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
897 tmp, prefix, |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
898 format->fill_char == '\0' ? ' ' : format->fill_char, |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
899 &locale, format->type == 'X'); |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
900 if (err) |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
901 Py_CLEAR(result); |
| 72475 | 902 |
| 903 done: | |
| 904 Py_XDECREF(tmp); | |
| 905 return result; | |
| 906 } | |
| 907 | |
| 908 /************************************************************************/ | |
| 909 /*********** float formatting *******************************************/ | |
| 910 /************************************************************************/ | |
| 911 | |
| 912 static PyObject* | |
| 913 strtounicode(char *charbuffer, Py_ssize_t len) | |
| 914 { | |
| 915 return PyUnicode_FromKindAndData(PyUnicode_1BYTE_KIND, charbuffer, len); | |
| 916 } | |
|
47393
0da27fdf0192
Refactor and clean up str.format() code (and helpers) in advance of optimizations.
Eric Smith <eric@trueblade.com>
parents:
42817
diff
changeset
|
917 |
| 72475 | 918 /* much of this is taken from unicodeobject.c */ |
| 919 static PyObject * | |
| 920 format_float_internal(PyObject *value, | |
| 921 const InternalFormatSpec *format) | |
| 922 { | |
| 923 char *buf = NULL; /* buffer returned from PyOS_double_to_string */ | |
| 924 Py_ssize_t n_digits; | |
| 925 Py_ssize_t n_remainder; | |
| 926 Py_ssize_t n_total; | |
| 927 int has_decimal; | |
| 928 double val; | |
| 929 Py_ssize_t precision = format->precision; | |
| 930 Py_ssize_t default_precision = 6; | |
| 931 Py_UCS4 type = format->type; | |
| 932 int add_pct = 0; | |
| 933 Py_ssize_t index; | |
| 934 NumberFieldWidths spec; | |
| 935 int flags = 0; | |
| 936 PyObject *result = NULL; | |
| 937 int maxchar = 127; | |
| 938 Py_UCS4 sign_char = '\0'; | |
| 939 int float_type; /* Used to see if we have a nan, inf, or regular float. */ | |
| 940 PyObject *unicode_tmp = NULL; | |
|
72495
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
941 int err; |
| 72475 | 942 |
| 943 /* Locale settings, either from the actual locale or | |
| 944 from a hard-code pseudo-locale */ | |
| 945 LocaleInfo locale; | |
| 946 | |
| 947 if (format->alternate) | |
| 948 flags |= Py_DTSF_ALT; | |
| 949 | |
| 950 if (type == '\0') { | |
| 951 /* Omitted type specifier. Behaves in the same way as repr(x) | |
| 952 and str(x) if no precision is given, else like 'g', but with | |
| 953 at least one digit after the decimal point. */ | |
| 954 flags |= Py_DTSF_ADD_DOT_0; | |
| 955 type = 'r'; | |
| 956 default_precision = 0; | |
| 957 } | |
| 958 | |
| 959 if (type == 'n') | |
| 960 /* 'n' is the same as 'g', except for the locale used to | |
| 961 format the result. We take care of that later. */ | |
| 962 type = 'g'; | |
| 963 | |
| 964 val = PyFloat_AsDouble(value); | |
| 965 if (val == -1.0 && PyErr_Occurred()) | |
| 966 goto done; | |
| 967 | |
| 968 if (type == '%') { | |
| 969 type = 'f'; | |
| 970 val *= 100; | |
| 971 add_pct = 1; | |
| 972 } | |
| 973 | |
| 974 if (precision < 0) | |
| 975 precision = default_precision; | |
| 976 else if (type == 'r') | |
| 977 type = 'g'; | |
| 978 | |
| 979 /* Cast "type", because if we're in unicode we need to pass a | |
| 980 8-bit char. This is safe, because we've restricted what "type" | |
| 981 can be. */ | |
| 982 buf = PyOS_double_to_string(val, (char)type, precision, flags, | |
| 983 &float_type); | |
| 984 if (buf == NULL) | |
| 985 goto done; | |
| 986 n_digits = strlen(buf); | |
| 987 | |
| 988 if (add_pct) { | |
| 989 /* We know that buf has a trailing zero (since we just called | |
| 990 strlen() on it), and we don't use that fact any more. So we | |
| 991 can just write over the trailing zero. */ | |
| 992 buf[n_digits] = '%'; | |
| 993 n_digits += 1; | |
| 994 } | |
| 995 | |
| 996 /* Since there is no unicode version of PyOS_double_to_string, | |
| 997 just use the 8 bit version and then convert to unicode. */ | |
| 998 unicode_tmp = strtounicode(buf, n_digits); | |
| 999 if (unicode_tmp == NULL) | |
| 1000 goto done; | |
| 1001 index = 0; | |
| 1002 | |
| 1003 /* Is a sign character present in the output? If so, remember it | |
| 1004 and skip it */ | |
| 1005 if (PyUnicode_READ_CHAR(unicode_tmp, index) == '-') { | |
| 1006 sign_char = '-'; | |
| 1007 ++index; | |
| 1008 --n_digits; | |
| 1009 } | |
| 1010 | |
| 1011 /* Determine if we have any "remainder" (after the digits, might include | |
| 1012 decimal or exponent or both (or neither)) */ | |
| 1013 parse_number(unicode_tmp, index, index + n_digits, &n_remainder, &has_decimal); | |
| 1014 | |
| 1015 /* Determine the grouping, separator, and decimal point, if any. */ | |
| 1016 get_locale_info(format->type == 'n' ? LT_CURRENT_LOCALE : | |
| 1017 (format->thousands_separators ? | |
| 1018 LT_DEFAULT_LOCALE : | |
| 1019 LT_NO_LOCALE), | |
| 1020 &locale); | |
| 1021 | |
| 1022 /* Calculate how much memory we'll need. */ | |
|
72495
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
1023 n_total = calc_number_widths(&spec, 0, sign_char, unicode_tmp, index, |
| 72475 | 1024 index + n_digits, n_remainder, has_decimal, |
| 1025 &locale, format); | |
| 1026 | |
| 1027 /* Allocate the memory. */ | |
| 1028 result = PyUnicode_New(n_total, maxchar); | |
| 1029 if (result == NULL) | |
| 1030 goto done; | |
| 1031 | |
| 1032 /* Populate the memory. */ | |
|
72495
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
1033 err = fill_number(result, 0, &spec, |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
1034 unicode_tmp, index, index + n_digits, |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
1035 NULL, 0, |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
1036 format->fill_char == '\0' ? ' ' : format->fill_char, |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
1037 &locale, 0); |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
1038 if (err) |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
1039 Py_CLEAR(result); |
| 72475 | 1040 |
| 1041 done: | |
| 1042 PyMem_Free(buf); | |
| 1043 Py_DECREF(unicode_tmp); | |
| 1044 return result; | |
| 1045 } | |
| 1046 | |
| 1047 /************************************************************************/ | |
| 1048 /*********** complex formatting *****************************************/ | |
| 1049 /************************************************************************/ | |
| 1050 | |
| 1051 static PyObject * | |
| 1052 format_complex_internal(PyObject *value, | |
| 1053 const InternalFormatSpec *format) | |
| 1054 { | |
| 1055 double re; | |
| 1056 double im; | |
| 1057 char *re_buf = NULL; /* buffer returned from PyOS_double_to_string */ | |
| 1058 char *im_buf = NULL; /* buffer returned from PyOS_double_to_string */ | |
| 1059 | |
| 1060 InternalFormatSpec tmp_format = *format; | |
| 1061 Py_ssize_t n_re_digits; | |
| 1062 Py_ssize_t n_im_digits; | |
| 1063 Py_ssize_t n_re_remainder; | |
| 1064 Py_ssize_t n_im_remainder; | |
| 1065 Py_ssize_t n_re_total; | |
| 1066 Py_ssize_t n_im_total; | |
| 1067 int re_has_decimal; | |
| 1068 int im_has_decimal; | |
| 1069 Py_ssize_t precision = format->precision; | |
| 1070 Py_ssize_t default_precision = 6; | |
| 1071 Py_UCS4 type = format->type; | |
| 1072 Py_ssize_t i_re; | |
| 1073 Py_ssize_t i_im; | |
| 1074 NumberFieldWidths re_spec; | |
| 1075 NumberFieldWidths im_spec; | |
| 1076 int flags = 0; | |
| 1077 PyObject *result = NULL; | |
| 1078 int maxchar = 127; | |
| 1079 int rkind; | |
| 1080 void *rdata; | |
| 1081 Py_ssize_t index; | |
| 1082 Py_UCS4 re_sign_char = '\0'; | |
| 1083 Py_UCS4 im_sign_char = '\0'; | |
| 1084 int re_float_type; /* Used to see if we have a nan, inf, or regular float. */ | |
| 1085 int im_float_type; | |
| 1086 int add_parens = 0; | |
| 1087 int skip_re = 0; | |
| 1088 Py_ssize_t lpad; | |
| 1089 Py_ssize_t rpad; | |
| 1090 Py_ssize_t total; | |
| 1091 PyObject *re_unicode_tmp = NULL; | |
| 1092 PyObject *im_unicode_tmp = NULL; | |
|
72495
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
1093 int err; |
| 72475 | 1094 |
| 1095 /* Locale settings, either from the actual locale or | |
| 1096 from a hard-code pseudo-locale */ | |
| 1097 LocaleInfo locale; | |
| 1098 | |
| 1099 /* Zero padding is not allowed. */ | |
| 1100 if (format->fill_char == '0') { | |
| 1101 PyErr_SetString(PyExc_ValueError, | |
| 1102 "Zero padding is not allowed in complex format " | |
| 1103 "specifier"); | |
| 1104 goto done; | |
| 1105 } | |
| 1106 | |
| 1107 /* Neither is '=' alignment . */ | |
| 1108 if (format->align == '=') { | |
| 1109 PyErr_SetString(PyExc_ValueError, | |
| 1110 "'=' alignment flag is not allowed in complex format " | |
| 1111 "specifier"); | |
| 1112 goto done; | |
| 1113 } | |
| 1114 | |
| 1115 re = PyComplex_RealAsDouble(value); | |
| 1116 if (re == -1.0 && PyErr_Occurred()) | |
| 1117 goto done; | |
| 1118 im = PyComplex_ImagAsDouble(value); | |
| 1119 if (im == -1.0 && PyErr_Occurred()) | |
| 1120 goto done; | |
| 1121 | |
| 1122 if (format->alternate) | |
| 1123 flags |= Py_DTSF_ALT; | |
| 1124 | |
| 1125 if (type == '\0') { | |
| 1126 /* Omitted type specifier. Should be like str(self). */ | |
| 1127 type = 'r'; | |
| 1128 default_precision = 0; | |
| 1129 if (re == 0.0 && copysign(1.0, re) == 1.0) | |
| 1130 skip_re = 1; | |
| 1131 else | |
| 1132 add_parens = 1; | |
| 1133 } | |
| 1134 | |
| 1135 if (type == 'n') | |
| 1136 /* 'n' is the same as 'g', except for the locale used to | |
| 1137 format the result. We take care of that later. */ | |
| 1138 type = 'g'; | |
| 1139 | |
| 1140 if (precision < 0) | |
| 1141 precision = default_precision; | |
| 1142 else if (type == 'r') | |
| 1143 type = 'g'; | |
| 1144 | |
| 1145 /* Cast "type", because if we're in unicode we need to pass a | |
| 1146 8-bit char. This is safe, because we've restricted what "type" | |
| 1147 can be. */ | |
| 1148 re_buf = PyOS_double_to_string(re, (char)type, precision, flags, | |
| 1149 &re_float_type); | |
| 1150 if (re_buf == NULL) | |
| 1151 goto done; | |
| 1152 im_buf = PyOS_double_to_string(im, (char)type, precision, flags, | |
| 1153 &im_float_type); | |
| 1154 if (im_buf == NULL) | |
| 1155 goto done; | |
| 1156 | |
| 1157 n_re_digits = strlen(re_buf); | |
| 1158 n_im_digits = strlen(im_buf); | |
| 1159 | |
| 1160 /* Since there is no unicode version of PyOS_double_to_string, | |
| 1161 just use the 8 bit version and then convert to unicode. */ | |
| 1162 re_unicode_tmp = strtounicode(re_buf, n_re_digits); | |
| 1163 if (re_unicode_tmp == NULL) | |
| 1164 goto done; | |
| 1165 i_re = 0; | |
| 1166 | |
| 1167 im_unicode_tmp = strtounicode(im_buf, n_im_digits); | |
| 1168 if (im_unicode_tmp == NULL) | |
| 1169 goto done; | |
| 1170 i_im = 0; | |
| 1171 | |
| 1172 /* Is a sign character present in the output? If so, remember it | |
| 1173 and skip it */ | |
| 1174 if (PyUnicode_READ_CHAR(re_unicode_tmp, i_re) == '-') { | |
| 1175 re_sign_char = '-'; | |
| 1176 ++i_re; | |
| 1177 --n_re_digits; | |
| 1178 } | |
| 1179 if (PyUnicode_READ_CHAR(im_unicode_tmp, i_im) == '-') { | |
| 1180 im_sign_char = '-'; | |
| 1181 ++i_im; | |
| 1182 --n_im_digits; | |
| 1183 } | |
| 1184 | |
| 1185 /* Determine if we have any "remainder" (after the digits, might include | |
| 1186 decimal or exponent or both (or neither)) */ | |
|
72495
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
1187 parse_number(re_unicode_tmp, i_re, i_re + n_re_digits, |
| 72475 | 1188 &n_re_remainder, &re_has_decimal); |
|
72495
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
1189 parse_number(im_unicode_tmp, i_im, i_im + n_im_digits, |
| 72475 | 1190 &n_im_remainder, &im_has_decimal); |
| 1191 | |
| 1192 /* Determine the grouping, separator, and decimal point, if any. */ | |
| 1193 get_locale_info(format->type == 'n' ? LT_CURRENT_LOCALE : | |
| 1194 (format->thousands_separators ? | |
| 1195 LT_DEFAULT_LOCALE : | |
| 1196 LT_NO_LOCALE), | |
| 1197 &locale); | |
| 1198 | |
| 1199 /* Turn off any padding. We'll do it later after we've composed | |
| 1200 the numbers without padding. */ | |
| 1201 tmp_format.fill_char = '\0'; | |
| 1202 tmp_format.align = '<'; | |
| 1203 tmp_format.width = -1; | |
| 1204 | |
| 1205 /* Calculate how much memory we'll need. */ | |
| 1206 n_re_total = calc_number_widths(&re_spec, 0, re_sign_char, re_unicode_tmp, | |
| 1207 i_re, i_re + n_re_digits, n_re_remainder, | |
| 1208 re_has_decimal, &locale, &tmp_format); | |
| 1209 | |
| 1210 /* Same formatting, but always include a sign, unless the real part is | |
| 1211 * going to be omitted, in which case we use whatever sign convention was | |
| 1212 * requested by the original format. */ | |
| 1213 if (!skip_re) | |
| 1214 tmp_format.sign = '+'; | |
| 1215 n_im_total = calc_number_widths(&im_spec, 0, im_sign_char, im_unicode_tmp, | |
| 1216 i_im, i_im + n_im_digits, n_im_remainder, | |
| 1217 im_has_decimal, &locale, &tmp_format); | |
| 1218 | |
| 1219 if (skip_re) | |
| 1220 n_re_total = 0; | |
| 1221 | |
| 1222 /* Add 1 for the 'j', and optionally 2 for parens. */ | |
| 1223 calc_padding(n_re_total + n_im_total + 1 + add_parens * 2, | |
| 1224 format->width, format->align, &lpad, &rpad, &total); | |
| 1225 | |
| 1226 result = PyUnicode_New(total, maxchar); | |
| 1227 if (result == NULL) | |
| 1228 goto done; | |
| 1229 rkind = PyUnicode_KIND(result); | |
| 1230 rdata = PyUnicode_DATA(result); | |
| 1231 | |
| 1232 /* Populate the memory. First, the padding. */ | |
| 1233 index = fill_padding(result, 0, | |
| 1234 n_re_total + n_im_total + 1 + add_parens * 2, | |
| 1235 format->fill_char=='\0' ? ' ' : format->fill_char, | |
| 1236 lpad, rpad); | |
| 1237 | |
| 1238 if (add_parens) | |
| 1239 PyUnicode_WRITE(rkind, rdata, index++, '('); | |
| 1240 | |
| 1241 if (!skip_re) { | |
|
72495
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
1242 err = fill_number(result, index, &re_spec, |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
1243 re_unicode_tmp, i_re, i_re + n_re_digits, |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
1244 NULL, 0, |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
1245 0, |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
1246 &locale, 0); |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
1247 if (err) { |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
1248 Py_CLEAR(result); |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
1249 goto done; |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
1250 } |
| 72475 | 1251 index += n_re_total; |
| 1252 } | |
|
72495
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
1253 err = fill_number(result, index, &im_spec, |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
1254 im_unicode_tmp, i_im, i_im + n_im_digits, |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
1255 NULL, 0, |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
1256 0, |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
1257 &locale, 0); |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
1258 if (err) { |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
1259 Py_CLEAR(result); |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
1260 goto done; |
|
cf3557b65ae1
fill_char() can now propagate an error
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72475
diff
changeset
|
1261 } |
| 72475 | 1262 index += n_im_total; |
| 1263 PyUnicode_WRITE(rkind, rdata, index++, 'j'); | |
| 1264 | |
| 1265 if (add_parens) | |
| 1266 PyUnicode_WRITE(rkind, rdata, index++, ')'); | |
| 1267 | |
| 1268 done: | |
| 1269 PyMem_Free(re_buf); | |
| 1270 PyMem_Free(im_buf); | |
| 1271 Py_XDECREF(re_unicode_tmp); | |
| 1272 Py_XDECREF(im_unicode_tmp); | |
| 1273 return result; | |
| 1274 } | |
| 1275 | |
| 1276 /************************************************************************/ | |
| 1277 /*********** built in formatters ****************************************/ | |
| 1278 /************************************************************************/ | |
| 1279 PyObject * | |
| 1280 _PyUnicode_FormatAdvanced(PyObject *obj, | |
| 1281 PyObject *format_spec, | |
| 1282 Py_ssize_t start, Py_ssize_t end) | |
| 1283 { | |
| 1284 InternalFormatSpec format; | |
|
72714
6f03716079a9
Don't check for the maximum character when copying from unicodeobject.c
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72705
diff
changeset
|
1285 PyObject *result; |
| 72475 | 1286 |
| 1287 /* check for the special case of zero length format spec, make | |
| 1288 it equivalent to str(obj) */ | |
|
72714
6f03716079a9
Don't check for the maximum character when copying from unicodeobject.c
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72705
diff
changeset
|
1289 if (start == end) |
|
6f03716079a9
Don't check for the maximum character when copying from unicodeobject.c
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72705
diff
changeset
|
1290 return PyObject_Str(obj); |
| 72475 | 1291 |
| 1292 /* parse the format_spec */ | |
| 1293 if (!parse_internal_render_format_spec(format_spec, start, end, | |
| 1294 &format, 's', '<')) | |
|
72714
6f03716079a9
Don't check for the maximum character when copying from unicodeobject.c
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72705
diff
changeset
|
1295 return NULL; |
| 72475 | 1296 |
| 1297 /* type conversion? */ | |
| 1298 switch (format.type) { | |
| 1299 case 's': | |
| 1300 /* no type conversion needed, already a string. do the formatting */ | |
| 1301 result = format_string_internal(obj, &format); | |
|
72714
6f03716079a9
Don't check for the maximum character when copying from unicodeobject.c
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72705
diff
changeset
|
1302 if (result != NULL) |
|
6f03716079a9
Don't check for the maximum character when copying from unicodeobject.c
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72705
diff
changeset
|
1303 assert(_PyUnicode_CheckConsistency(result, 1)); |
| 72475 | 1304 break; |
| 1305 default: | |
| 1306 /* unknown */ | |
| 1307 unknown_presentation_type(format.type, obj->ob_type->tp_name); | |
|
72714
6f03716079a9
Don't check for the maximum character when copying from unicodeobject.c
Victor Stinner <victor.stinner@haypocalc.com>
parents:
72705
diff
changeset
|
1308 result = NULL; |
| 72475 | 1309 } |
| 1310 return result; | |
| 1311 } | |
| 1312 | |
| 1313 static PyObject* | |
| 1314 format_int_or_long(PyObject* obj, PyObject* format_spec, | |
| 1315 Py_ssize_t start, Py_ssize_t end, | |
| 1316 IntOrLongToString tostring) | |
| 1317 { | |
| 1318 PyObject *result = NULL; | |
| 1319 PyObject *tmp = NULL; | |
| 1320 InternalFormatSpec format; | |
| 1321 | |
| 1322 /* check for the special case of zero length format spec, make | |
| 1323 it equivalent to str(obj) */ | |
| 1324 if (start == end) { | |
| 1325 result = PyObject_Str(obj); | |
| 1326 goto done; | |
| 1327 } | |
| 1328 | |
| 1329 /* parse the format_spec */ | |
| 1330 if (!parse_internal_render_format_spec(format_spec, start, end, | |
| 1331 &format, 'd', '>')) | |
| 1332 goto done; | |
| 1333 | |
| 1334 /* type conversion? */ | |
| 1335 switch (format.type) { | |
| 1336 case 'b': | |
| 1337 case 'c': | |
| 1338 case 'd': | |
| 1339 case 'o': | |
| 1340 case 'x': | |
| 1341 case 'X': | |
| 1342 case 'n': | |
| 1343 /* no type conversion needed, already an int (or long). do | |
| 1344 the formatting */ | |
| 1345 result = format_int_or_long_internal(obj, &format, tostring); | |
| 1346 break; | |
| 1347 | |
| 1348 case 'e': | |
| 1349 case 'E': | |
| 1350 case 'f': | |
| 1351 case 'F': | |
| 1352 case 'g': | |
| 1353 case 'G': | |
| 1354 case '%': | |
| 1355 /* convert to float */ | |
| 1356 tmp = PyNumber_Float(obj); | |
| 1357 if (tmp == NULL) | |
| 1358 goto done; | |
| 1359 result = format_float_internal(tmp, &format); | |
| 1360 break; | |
| 1361 | |
| 1362 default: | |
| 1363 /* unknown */ | |
| 1364 unknown_presentation_type(format.type, obj->ob_type->tp_name); | |
| 1365 goto done; | |
| 1366 } | |
| 1367 | |
| 1368 done: | |
| 1369 Py_XDECREF(tmp); | |
| 1370 return result; | |
| 1371 } | |
| 1372 | |
| 1373 /* Need to define long_format as a function that will convert a long | |
| 1374 to a string. In 3.0, _PyLong_Format has the correct signature. */ | |
| 1375 #define long_format _PyLong_Format | |
| 1376 | |
| 1377 PyObject * | |
| 1378 _PyLong_FormatAdvanced(PyObject *obj, | |
| 1379 PyObject *format_spec, | |
| 1380 Py_ssize_t start, Py_ssize_t end) | |
| 1381 { | |
| 1382 return format_int_or_long(obj, format_spec, start, end, | |
| 1383 long_format); | |
| 1384 } | |
| 1385 | |
| 1386 PyObject * | |
| 1387 _PyFloat_FormatAdvanced(PyObject *obj, | |
| 1388 PyObject *format_spec, | |
| 1389 Py_ssize_t start, Py_ssize_t end) | |
| 1390 { | |
| 1391 PyObject *result = NULL; | |
| 1392 InternalFormatSpec format; | |
| 1393 | |
| 1394 /* check for the special case of zero length format spec, make | |
| 1395 it equivalent to str(obj) */ | |
| 1396 if (start == end) { | |
| 1397 result = PyObject_Str(obj); | |
| 1398 goto done; | |
| 1399 } | |
| 1400 | |
| 1401 /* parse the format_spec */ | |
| 1402 if (!parse_internal_render_format_spec(format_spec, start, end, | |
| 1403 &format, '\0', '>')) | |
| 1404 goto done; | |
| 1405 | |
| 1406 /* type conversion? */ | |
| 1407 switch (format.type) { | |
| 1408 case '\0': /* No format code: like 'g', but with at least one decimal. */ | |
| 1409 case 'e': | |
| 1410 case 'E': | |
| 1411 case 'f': | |
| 1412 case 'F': | |
| 1413 case 'g': | |
| 1414 case 'G': | |
| 1415 case 'n': | |
| 1416 case '%': | |
| 1417 /* no conversion, already a float. do the formatting */ | |
| 1418 result = format_float_internal(obj, &format); | |
| 1419 break; | |
| 1420 | |
| 1421 default: | |
| 1422 /* unknown */ | |
| 1423 unknown_presentation_type(format.type, obj->ob_type->tp_name); | |
| 1424 goto done; | |
| 1425 } | |
| 1426 | |
| 1427 done: | |
| 1428 return result; | |
| 1429 } | |
| 1430 | |
| 1431 PyObject * | |
| 1432 _PyComplex_FormatAdvanced(PyObject *obj, | |
| 1433 PyObject *format_spec, | |
| 1434 Py_ssize_t start, Py_ssize_t end) | |
| 1435 { | |
| 1436 PyObject *result = NULL; | |
| 1437 InternalFormatSpec format; | |
| 1438 | |
| 1439 /* check for the special case of zero length format spec, make | |
| 1440 it equivalent to str(obj) */ | |
| 1441 if (start == end) { | |
| 1442 result = PyObject_Str(obj); | |
| 1443 goto done; | |
| 1444 } | |
| 1445 | |
| 1446 /* parse the format_spec */ | |
| 1447 if (!parse_internal_render_format_spec(format_spec, start, end, | |
| 1448 &format, '\0', '>')) | |
| 1449 goto done; | |
| 1450 | |
| 1451 /* type conversion? */ | |
| 1452 switch (format.type) { | |
| 1453 case '\0': /* No format code: like 'g', but with at least one decimal. */ | |
| 1454 case 'e': | |
| 1455 case 'E': | |
| 1456 case 'f': | |
| 1457 case 'F': | |
| 1458 case 'g': | |
| 1459 case 'G': | |
| 1460 case 'n': | |
| 1461 /* no conversion, already a complex. do the formatting */ | |
| 1462 result = format_complex_internal(obj, &format); | |
| 1463 break; | |
| 1464 | |
| 1465 default: | |
| 1466 /* unknown */ | |
| 1467 unknown_presentation_type(format.type, obj->ob_type->tp_name); | |
| 1468 goto done; | |
| 1469 } | |
| 1470 | |
| 1471 done: | |
| 1472 return result; | |
| 1473 } |